Cursor releases Composer 2, a frontier coding model at $0.50 per million tokens

Cursor has unveiled Composer 2, a new coding model that delivers frontier-level performance and changes the economics of AI coding. The model is available now in the Cursor editor and comes with a technical report on its training.

Frontier-level results on benchmarks

Composer 2 improves substantially on all the main coding benchmarks. On Cursor's own CursorBench 2, the model scored 61.3, which is 38% better than Composer 1.5. The results are even more striking on SWE-bench Multilingual, where Composer 2 scored 73.7 against 65.9 for the previous version. On Terminal-Bench 2.0, it scored 61.7.

• CursorBench: 61.3 (+38% vs Composer 1.5)
• Terminal-Bench 2.0: 61.7
• SWE-bench Multilingual: 73.7 (+13% vs Composer 1.5)

The gains come from the first round of continued pretraining, in which the model is retrained on improved data. This gives a stronger base for the reinforcement learning that follows, which teaches the model to solve complex multi-step tasks.

A price that changes the rules

Composer 2 costs $0.50 per million input tokens and $2.50 per million output tokens. For comparison, it is one of the most affordable options on the market at this level of quality.

Cursor has also released a faster version of the model with the same performance, priced at $1.50 for input and $7.50 for output tokens. The fast variant is now the default, and its price is lower than that of competitors' fast models.

This combination of high capability and low price makes Composer 2 the best choice for many development workflows. On individual plans, Composer usage falls into a separate pool with a generous limit.

How it was done

The improvements rest on a new training approach. Cursor first ran continued pretraining on better coding data, creating a stronger starting point for further optimization. The model was then trained on long-horizon tasks through reinforcement learning, meaning tasks that require hundreds of actions and decisions. This is what lets Composer 2 handle truly complex coding scenarios, including multi-file projects and architectural decisions.

What it means

The AI coding market is entering a new period in which quality and cost no longer have to be traded off against each other. For developers, this means powerful coding tools are becoming more practical and affordable. For companies like Cursor, it signals that the line between standard and frontier now runs along training speed and cost efficiency, not along a model's absolute performance.